Date: Fri, 27 Dec 1996 15:34:40 -0700 (MST) From: Terry Lambert <terry@lambert.org> To: joerg_wunsch@uriah.heep.sax.de Cc: freebsd-hackers@FreeBSD.org Subject: Re: unlink by inodes? Message-ID: <199612272234.PAA25350@phaeton.artisoft.com> In-Reply-To: <199612191917.UAA28344@uriah.heep.sax.de> from "J Wunsch" at Dec 19, 96 08:17:12 pm
next in thread | previous in thread | raw e-mail | index | archive | help
J"org writes: > As Chris Csanady wrote: > > > Yeah. :) You really should try to use the fs as it was intended > > anyway. BTW, does anyone know the status of McKusick's soft > > updates integration? I think this would make a lot of people happy.. > > Do you perchance mean `mount -o async'? No, I don't think he does. Someone at the CSRG was rumored to be working on yet another 4.4BSD release. Rumor was that it was to include the Soft Updates code from the WWW posted Appendix A to the Ganger/Patt paper on Metadata updates. The Ganger/Patt code is from SVR4 and has been "sanitized" (read: SVR4 has mostly been ripped out, and a number of coding errors have been introduced). The problem with the Ganger/Patt code (besides that fact that Matt Day, mday@elbereth.org, has already integrated it last year on top of my layering patches required to make the namespace code work on the Artisoft Windows95 IFS port of the BSD4.4 Heidemann framework) is that it is not generalized... that is, it would be hard to maintain LFS, MFS, and so on in the face of a hacked up UFS layer. I am really skeptical of any soft updates implementation that doesn't approach a file system as an event/action based system with commutative and associative properties for the event/action node relations. Doing anything less means that you could not unify the model and abstract it from a particular implementation (in this case, SVR4 UFS or BSD UFS and FFS layers). My personal opinion is that any implementation short of this would end up being a kludge that would live on in infamy, a historical wart we could never successfully kill, for ever after. I've talked to Ganger a bit on this, but he isn't very hip on graphical least-path soloutions, and so wasn't terribly interested (all his work is in SVR4 anyway, so he doesn't have to deal with layer interactions between stacking layers anyway, so for him it's a non-problem). Regards, Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199612272234.PAA25350>